Skip to content

sdbds/ACE-Step-1.5-for-windows

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1,425 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Ace-Step-1.5-for-windows

original backend codebase from ACE-Step-1.5

https://github.com/ace-step/ACE-Step-1.5

original frontend codebase from ace-step-ui

https://github.com/fspecii/ace-step-ui

image

Feather

1. Complete style search and random selection, with a total of 936 styles, synchronized from Suno's explorer.

image

2.Generate a complete record of song parameters, which can be reused at any time.

image

3.All pages are localized into four languages ​​(English, Chinese, Japanese, and Korean).

image

4.LoRA training support and memory offloading optimization.

There were still some issues with the original backend repository's GPU memory unloading, so we made significant reconstructions and optimizations. Now we train with the lowest GPU memory and the fastest speed.

image

5.LoKR training support and reduce training time and improve sound quality

image

6.We've also added the ability to load Lora/LoKR. Note that Lora reads folders, while LoKR reads safetensors files.

image

🔧 Setting up the Environment for Windows

Give unrestricted script access to powershell so venv can work:

  • Open an administrator powershell window
  • Type Set-ExecutionPolicy Unrestricted and answer A
  • Close admin powershell window

Installation

Clone the repo with --recurse-submodules:

git clone --recurse-submodules https://github.com/sdbds/ACE-Step-1.5-for-windows.git -b qinglong

MUST USE --recurse-submodules

Change Default Model

copy .env.sample and rename to .env then change model name for which your choose.

Windows

Run the following PowerShell script:

./1install-uv-qinglong.ps1

(Optional)

VS Studio 2022 for torch compile

Download from Microsoft offical link: https://aka.ms/vs/17/release/vs_community.exe

Install C++ desktop and language package with English(especially for asian computer)

FFMPEG

https://github.com/BtbN/FFmpeg-Builds/releases/download/latest/ffmpeg-n8.0-latest-win64-gpl-shared-8.0.zip

use shared version for ffmpeg

Linux

  1. First install PowerShell:
./0、install pwsh.sh
  1. Then run the installation script using PowerShell:
sudo pwsh ./1install-uv-qinglong.ps1

use sudo pwsh if you in Linux without root user.

Usage

Run

3、run_server.ps1

for API_backend

4、run_npmgui.ps1

for npm_frontend

About

The most powerful local music generation model that outperforms most commercial alternatives

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages

  • Python 90.8%
  • Shell 4.5%
  • Batchfile 3.1%
  • HTML 0.7%
  • JavaScript 0.6%
  • TypeScript 0.3%